remittances, public health spending and foreign aid in the

HAL Id: halshs-00552996https://halshs.archives-ouvertes.fr/halshs-00552996

Preprint submitted on 6 Jan 2011

HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, estdestinée au dépôt et à la diffusion de documentsscientifiques de niveau recherche, publiés ou non,émanant des établissements d’enseignement et derecherche français ou étrangers, des laboratoirespublics ou privés.

Remittances, Public Health Spending and Foreign Aidin the Access to Health Care Services in Developing

CountriesAlassane Drabo, Christian Hubert Ebeke

To cite this version:Alassane Drabo, Christian Hubert Ebeke. Remittances, Public Health Spending and Foreign Aid inthe Access to Health Care Services in Developing Countries. 2011. �halshs-00552996�

https://halshs.archives-ouvertes.fr/halshs-00552996

https://hal.archives-ouvertes.fr

CERDI, Etudes et Documents, E 2010.04

1

Document de travail de la série

Etudes et Documents

E 2010.04

Remittances, Public Health Spending

and Foreign Aid in the Access to Health Care

Services in Developing Countries

Alassane Drabo and Christian Ebeke 1

1Centre d’Etudes et de Recherches sur le Développement International (CERDI-CNRS), Université d’Auvergne. 65 Bd François

Mitterrand 63000 Clermont Ferrand, France. Tel: (+33) 4 73 17 75 07 Fax: (+33) 4 73 17 74 28. e-mail: Alassane.Drabo@u-

clermont1.fr (Alassane Drabo), [email protected] (Christian Ebeke)

We would like to thank Bernadette Kamgnia Dia, Audibert Martine, Mathonnat Jacky, Huang Xiaoxian, Hélène Ehrhart and Emilie

Caldeira and all the participants to the 11th

annual Conference of the Global Development Network in Prague for their helpful

comments and discussions. We are also grateful to participants at the CERDI doctoral seminar held in January 2010.


2

Abstract

The aim of this paper is to analyze the respective impacts of remittances, health aid and public spending on the access to health care services in developing countries. The specific objectives are threefold. Firstly, we quantify the differential impacts of remittances on the access to public and private health care services. Secondly, we determine whether remittances and foreign health aid are complements or substitutes in the access to health care services. Lastly, we evaluate the heterogeneity of the impact of remittances in the access to public and private health care services by quintile of income. We provide a rigorous econometric analysis by controlling for the endogeneity of remittances, public spending and foreign aid. We find that remittances, health aid and public spending are important determinants of access to health services in recipients’ countries. Another interesting result comes from the fact that, remittances lead to a sectorial glide in the uses of health care services from the public to the private sector for the intermediate income class. This result holds also for the richer quintiles that are the major recipients of remittances in developing countries. Moreover, remittances and foreign health aid are complements for the access to health care services in “low” income countries. Finally, these results suggest that policies aiming at increasing remittances are appropriate for developing countries but also that, the “optimal” therapy for the “low” income countries is the combination of remittances and foreign aid.

Keywords: Remittances, health aid, public spending, access to health care services, developing countries, instrumental variables method

JEL codes: I1 F24 O19


3

1. INTRODUCTION

Several arguments can be evoked in explaining the interest on studies undertaken on health care in developing countries. First of all, the status of health of the populations is a major preoccupation insofar as it determines the level of productivity of the labor force and contributes to growth as well as to poverty reduction (Bloom and al., 2004; Carstensen & Gundlach, 2006; Weil, 2007). The key role of health as input for development has been reaffirmed at the international level, as evidenced by the Millennium Development Goals (MDGs). Indeed, three of the eight objectives of the MDGs are centered directly on health. These are: the maternal health, the fight against HIV AIDS and the reduction of child mortality. Secondly, the paradox to which developing countries are confronted rest on the size of their needs in the access to health care services in a context of severe financial constraints. Public aid to development that remains one of the main sources of the external financing of health care services in developing countries in general and in Africa in particular is rather stagnating whereas the possibilities of the internal sources of financing health care in these countries are further constrained by the lack of fiscal resources and by numerous institutional problems.

Indeed, a number of recent studies on health care systems in Africa pointed out the existence of substantial gaps between budgets voted at the central level and disposable resources that are actually received by the local communities (Gauthier & Wayne, 2009; Reinikka & Svensson, 2004, 2005). Beside the problem of corruption and bad governance (Rajkumar & Swaroop, 2008), the concerns for the progressivity of the advantages of public spending on health or more generally, the question of equity in the access to health care services remain acute in the context of developing countries in general, and of African countries in particular (Castro - Leal and al., 1999; Gwatkin, 2003; Gwatkin and al., 2007; Kamgnia, 2008; Kamgnia and al., 2008; Berthelemy & Seban, 2009; Yazbeck, 2009).

If public spending and foreign aid on health stand as key sources of financing health and of inequality reduction in the access to health care services in developing countries, other important sources for financing heath care services do exist as well. Indeed, several recent papers (Edward & Uretas, 2003, Chauvet and al., 2008, Amuedo–Dorantes and al., 2007) pointed out the important role that remittances play in the financing of households’ health. Remittances constitute one of the most important features of the current globalization and Ratha & Zhimey (2008) for the World Bank estimate to 240 billion of US dollars, the total volume of funds received by developing countries through these channels in 2007.

The positive impact of remittances on development led to a number of studies. The positive effects of remittances on the accumulation of human capital (Borraz, 2005; Hanson & Woodruff, 2003; Calero and al., 2008; Yang, 2008; Bansak & Chezum, 2009), on the accumulation of physical capital (Woodruff & Zenteno, 2007), on the reduction of income inequalities (Koechlin & Léon, 2007; Chauvet & Mesplé-Somps, 2007), on consumption (Adams and al., 2008), on economic growth (Giuliano & Ruiz - Arranz, 2008; Catrinescu and al., 2009) and on poverty reduction (Adams & Page, 2005; Gupta and al., 2009) have been well demonstrated.

The interest for studies on remittances is justified on at least three grounds. Firstly, remittances differ from the other types of funds received by developing countries insofar as they directly go in the pocket of the households. Secondly, remittances exhibit a relative stability, or to say a somehow countercyclical evolution, contrary to the other sources of funds received by developing countries. Thirdly, following the international statistical databases, the volume of remittances flowing into developing countries exceeds the flows of aid, what makes remittances a non negligible source of funds of which it becomes imperative to size all the effects on receiving countries.

The current study expands on existing empirical studies that have analyzed the impact of remittances on the accumulation of human capital. However it differs from the other studies on that it stresses on the causal effects of remittances on the level of access to health care services as well as on equity in the access to health services.


4

The general objective of the current study is to analyze the contributions of remittances, public spending and foreign aid in the access to health care services in developing countries. The specific objectives are:

� to quantify the differential impacts of remittances in the access of public and private health care services;

� to determine the interactions between remittances and foreign health aid in the overall access to health care services;

� to evaluate the heterogeneity of the impact of remittances in the access to public and private health care services by quintile of income.

These hypotheses are tested by using instrumental variables approaches, a method which allows us to deal with the potential endogeneity of our variables of interest. Using a comprehensive and recent database on intra-country access to health care services at the international level (Health, Nutrition and Population (HNP) database), our results show that remittances, health aid and public spending are important factors of access to health services in developing countries. Moreover, we have found that migrants’ remittances appear to be complementary to health aid in the “low” income countries and not necessarily in the others. We also find that remittances lead to a sectorial glide in the uses of health care services from the public to the private sector for the intermediate income class and for the richer quintiles that are the major recipients of remittances in the developing world.

The rest of the paper is organized as follow. Section 2 is devoted to the literature review of the determinants of access to health care services in developing countries. Section 3 presents firstly a simple theoretical model of the relationship between remittances and the demand for medical services in a constrained environment, and secondly the econometric equations, estimation method and data used. We discuss our econometric results in section 5 while section 6 concludes on policy implications.

2. LITERATURE REVIEW

2.1. Traditional determinant of access to health services in developing countries

Bryant et al. (2002) studied the extent to which selected social and economic factors are influential in determining the participation in preventive cancer screening programs (namely screening mammography and Pap tests) in Prince George, a large community in northern British Columbia. Using the 1994 National Population Health Survey, which contains a sample of 416 women, they show through a series of logistic regression analyses that immigrant women, single women and women with less education in northern British Columbia experience low participation in health services, resulting in a higher risk for poor health and a poor quality of life. They also found no association between social factors and previous mammography use.

Falkingham (2004) uses the Tajikistan Livings Standard Survey to examine the extent to which the level and the distribution of out-of-pocket payments for health care payments act as barriers to health-care access. He shows that there are significant differences in health-care utilization rates across socio-economic groups and that these differences are related to ability to pay.

Using multilevel logistic regression models and data from the 2001 Demographic and Health Survey of Mali, Gage (2007) explores factors that determine the utilization of maternal health care in rural Mali. She shows that the dearth of health facilities, the lack of transportation means, distance, household poverty and personal problems are some barriers to the access of maternal health care services. In addition, living in area where women visit health services and living in close proximity to people with secondary or higher education increase the probability of women to use maternal health services.


5

Sepehri and al. (2008) examine the impacts of individual, household and commune-level characteristics on a woman’s decision to seek prenatal care, on the number of prenatal visits and on the choice between giving birth at a health facility or at home. They used data from the Vietnam latest National Household Survey data for 2001–2002 and a random intercept logistic model specification. They show that women decision regarding the place of child birth and the utilization of prenatal care are jointly determined by the observed individual characteristics (insurance status, level of education, marital status, birth order of the child, unwanted pregnancy), household characteristics (income class, household size, ethnicity, distance to the nearest hospital) and commune-level characteristics (urban or rural, income per capita) as well as by the unobserved commune-level characteristics.

Murawski & Church (2009) propose a model that can be used to maximize the access to health facilities by making selected improvements to the road system. They have applied their model to the Suhum District of Ghana and their application shows that significant increases to all season health services access can be obtained by a modest level of road investment.

Using household survey and ministry data in some African countries (South Africa, Cote d’Ivoire, Ghana, Guinea, Kenya, Madagascar and Tanzania), Castro-Leal and al. (2000) show that public spending in health favours mostly the better-off rather than the poor.

Seban & Berthelemy (2009) is one of the first papers using macroeconomic data to address the issue of the access to health services in developing countries. Using Health, Nutrition and Population (HNP) data compiled by Gwatkin (2007), they show that access to health services depends on some socioeconomic characteristics of household such as mother’s education level and wealth, but they find no evidence that public health expenditure is a robust determinant of access to health services in developing countries. However, one of the main drawback of their empirical analysis is they have not taken into account the potential endogeneity of public spending on health, what can be lead to severe bias in the parameters estimated.

2.2. International migration and remittances in the literature on health in developing countries

The literature on the link between international migration and health in developing countries is recent and can be split into two broad research areas. On the one hand, authors have analyzed the impact of international migration and remittances directly on the health status in recipient countries. On the other hand, the question turns on the impact of migration and remittances on the access to health care.

International migration, remittances and health status

Few papers have addressed the impact of migration and remittances on health. The exceptions are mostly focused on infant mortality. Brockerhoff (1990) and Ssengonzi and al. (2002) investigate the effects of female migration on the survival chances of their children in Senegal and Uganda, respectively. They find that rural to urban migration significantly increases child survival chances. With data on Mexican municipalities, Lopez-Cordova (2004) concludes that larger proportions of remittances and migrant households at the community level are associated with lower infant mortality rates.

Kanaiaupuni and Donato (1999) analyzed the effects of village migration and remittances on infant survival outcomes in Mexico, and concluded that remittances reduce infant mortality. However, the authors reach an opposite conclusion for the effect of migration: higher rates of infant mortality in communities experiencing intense migration.

Further evidence of the impact of migration on child health has been provided by Hildebrandt and McKenzie (2005). The authors investigate the impact of international migration on several children health outcomes in Mexico. Their results show that migrant households have lower rates of infant mortality and higher birth weights. Moreover, they find evidence that migration also raises maternal health knowledge and the likelihood that the child was delivered by a doctor. On the other hand, preventative health care (such as


6

breastfeeding, visits to doctors, and vaccinations) seem to be less likely for children from migrant households.

Fajnzylber and Lopez (2007) analyze the effect of international migrants’ remittances on anthropometric child health indicators in two countries, Guatemala and Nicaragua. They also estimate the impact of remittances on the probability that the delivery of children born in the year preceding the survey was assisted by a doctor, and on the probability that children aged 2 to 5 received the complete set of required vaccinations. The anthropometric measures on which they focus are the Weight-for- Age (WAZ) and Height-for-Age (HAZ) z-scores for children aged 1 to 5 years old. Their results suggest that children from households that report receiving remittances tend to exhibit higher health outcomes than those from non-recipients households with similar demographic and socio-economic characteristics. The results also indicate that the impact of remittances on children health is concentrated on low income households located in the first quintile of the income distribution.

Bhargava and Docquier (2008) use Country-level longitudinal data at three-year intervals over 1990–2004 to analyze the factors affecting emigration of physicians from Sub-Saharan countries and the effects of this medical brain drain on life expectancy and number of deaths due to AIDS. They estimate a triangular system of equations in a random effects specification for medical brain drain rates, life expectancy, and numbers of deaths due to AIDS. They found that in countries in which the HIV prevalence rate exceeds 3 percent, a doubling of the medical brain drain rate is associated with a 20 percent increase in adult deaths from AIDS.

Chauvet and al. (2008) analyze the respective impact of aid and remittances on human development as measured by infant and child mortality rates with a panel data on a sample of 109 developing countries, and cross-country quintile-level data on a sample of 47 developing countries. Their results suggest that remittances significantly improve child health. By contrast, medical brain drain, as measured by the expatriation rate of physicians, is found to have a harmful impact on health outcomes, so net impact of migration on human development is therefore mitigated. Lastly, remittances seem to be much more effective in improving health outcomes for children belonging to the richest households.

Migration, remittances and access to health services

Increased spending on health services by migrant households may improve the access to private sector health services for everyone in the community as service providers move to the community in response to this increased demand. In a study of the consequences of migration in Kerala (India), Zachariah and al. (2001) find that the receipt of remittances is associated with greater use of hospital facilities to treat illness, and a three-fold increase in the use of private hospitals for childbirth.

Lindstrom & Munoz-Franco (2006) use data from the 1995 Guatemalan Survey of Family Health (EGSF) conducted in four rural regions of Guatemala to explore how migration experience and social ties to migrants influence the likelihood of using formal maternal health-care services. They identify assimilation, diffusion, and remittances as three potential pathways through which migration can affect health-care services utilization in rural areas. They find that urban migration experience and having relatives abroad, are associated with a greater likelihood of formal prenatal care utilization, after taking account of background characteristics and enabling resources. Migration experience at all levels is also strongly associated with formal delivery assistance; however, this association operates primarily through the positive association between migration and enabling resources. Moreover, financial cost and geographic access are the most important barriers to formal delivery assistance, whereas awareness and acceptance remain as important barriers to the use of formal prenatal care in rural Guatemala.

Reanne et al. (2009) use data from a 2006 survey of two localities in the municipal city of Tepoztlán and Morelos (Mexico). They used logistic regressions to determine whether household remittances expenditures on health care were conditional to the type of health insurance coverage. They found that individuals who lacked insurance coverage or who were covered by the Seguro Popular program are significantly more likely


7

to reside in households that spend remittances on health care than individuals covered by an employer-based insurance program.

3. THEORETICAL MODEL AND EMPIRICAL SPECIFICATIONS

The three following hypotheses are successively tested in this paper:

� remittances increase the access to private health care services more than that of public services in developing countries while foreign aid and public spending increase more the utilization of public health care services;

� remittances and foreign aid are complements in poor countries and substitutes in rich countries in the access to health services.

� remittances lead to a sectorial glide in the uses of health care services from the public to the private sectors for the intermediate and the richer income classes that are the major recipients of remittances in developing countries.

3.1. Theoretical background: A model of access to medical services in presence of remittances

The model presented here is a simplified version of Grossman (1972) model. We exploit the specification of Zweifel et al. (1997), but we augment the model with the possibility that the individual can receive remittances to finance health expenditures. Consider an individual with a planning horizon of two time periods. During each period he or she

experiences a nonnegative amount of sick time s

t , which is the lower the larger health stock H. The individual derives a positive utility from consumption goods X while deriving disutility from sick time

( )Ht s. The utility function defined over these arguments is assumed to be time independent. Future utility is

discounted by a subjective factor β ≤ 1. Thus, the individual maximizes discounted utility U,

( )( ) ( )( )1100 ,, XHtUXHtUUss β+= with 0<

∂

∂s

t

U,

( )0

2

2

>∂

∂s

t

U, ,0>∂

∂

X

U,0

2

2

<∂

∂

X

U.0<

∂

∂

H

ts

(1)

The crucial component of the Grossman model is the equation that defines the change in the health stock

over time. On the one hand, health capital depreciates at a rate δ , causing health to worsen over time. On the other hand, the individual can increase health capital by investing I. This involves the purchase of

medical services M or spending I

t units of time on preventive effort. In all, one has:

( ) ( )ItMIHH ,1 001 +−= δ with ,0>

∂

∂

M

I,0>

∂

∂I

t

I,0

2

2

<∂

∂

M

I

( ).0

2

2

<∂

∂I

t

I (2)

Equation (2) constitutes a constraint that will enter the individual’s maximization problem.


8

The second constraint faced by the individual is the budget constraint. Our innovation here is that, abstracting from health insurance, health care expenditure amounting to pM has to be financed out of labor

income, initial wealth 0A or remittances received from abroad L, with w0 denoting the wage rate of the

initial period and p denoting the price of medical care. Remittances are targeted to buy health care, more

precisely, to cover a fraction φ ( )10 ≤< φ of health care expenditures pM.

Consumption (at price c) must be positive in both periods. Total time available is normalized at 1 in both periods. Altogether, the following budget constraint holds after discounting to present value with the return of saving R,

( ) ( )( ) ( ) 10111000

111

11 cX

RcXpMHtw

RttwA

sIs++−=−+−−+ φ (3)

To solve this maximization problem, the Lagrangean method is used,

( ) ( )( ) ( )( ) ( ) ( )( )1001100101 ,1,,,,,, HtMIHXHtUXHtUXXMtHIssI −+−++= δµβl

( ) ( )( ) ( )

−−−−−+−−++ 10111000

111

11 cX

RcXpMHtw

RttwA

sIsφλ (4)

First-order conditions for an optimum are given by the derivatives with respect to all decisional variables. We obtain:

01

1

11

=−∂

∂−

∂

∂

∂

∂=

∂

∂µ

λβ

H

tw

RH

t

t

U

H

ss

s

l (5)

00 =−∂

∂=

∂

∂w

t

I

tII

λµl

(6)

( ) 01 =−−∂

∂=

∂

∂φλµ p

M

I

M

l (7)

000

=−∂

∂=

∂

∂c

X

U

Xλ

l (8)

.011

=−∂

∂=

∂

∂c

RX

U

X

λβ

l (9)


9

Firstly, dividing equation (6) by equation (7) yields ( )φ−

=

∂∂

∂∂

10

p

w

MI

tI

I

(10)

Secondly, dividing equation (8) by equation (9) yields R

XU

XU

β=

∂∂

∂∂

1

0 (11)

Finally, equation (5) may be rewritten by solving (9) for R

λ,

µβ =

∂

∂−

∂

∂

∂

∂−

s

s

t

U

X

U

c

w

H

t

1

1

1

( )φλ −

∂∂

= 1

MI

p

( )φ−

∂∂

∂∂

= 10

c

p

MI

XU

, using (7) and (8) (12)

Our principal objective is to derive an explicit equation of medical services demand. To achieve this, we

need to specify the investment function ( )ItMI , . It is assumed to be of the Cobb-Douglas type,

( ) ,1 EI E

MM etMI

ααα −= ,10 << Mα 0>Eα . (13)

Note that education E serves to magnify the effects of medical services M and time spent in favour of health

.It

The structural demand for medical services

To derive the structural demand curve for medical services, we first logarithmize the investment function

(13). This leads to ( ) EtMI E

I

MM ααα +−+= ln1lnln (14)

With www == 10 , equation (10) can thus be rewritten as: ( )

( )φα

α

−=

−

1

1

p

w

t

M

M

M

I (15)

Taking logarithms results in ( ) ( ) ( ) ( )

−+−+−+=

M

MIwMpt

α

αφ

1lnlnln1lnlnln (16)

By substituting this expression into (14) yields the following equation for Mln in the case where ,1=δ

and thus, 1lnln HI = by equation (2),


10

( ) ( ) ( ) ( ) ( ) ( )

−+−−+−−−−−=

M

MEMMM EwpHM

α

αααφαα

1lnln11ln1ln1lnln 1 (17)

This expression is the equation of structural demand for medical services. We can see easily that, the higher the amount of remittances received, the higher the quantity demanded. The reduced demand function for medical services

The structural demand function for medical services (17) depends “unfortunately” on the health capital 1H

which is chosen optimally by the individual. To solve this problem, one needs first, to specify how sick time s

t depends on health capital 1H . The following functional form is assumed,

( ) ,2

111

θθ

−= HHt s

,01 >θ 02 >θ . (18)

The derivative of s

t with respect to 1H gives, 1

121

1

2 −−−=

∂

∂ θθθ H

H

ts

(19)

In the pure investment model, 0=∂

∂s

t

U. Thus, equation (12) becomes,

( )φβ −

∂∂

∂∂

=

∂

∂

∂

∂− 10

1

1

1 c

p

MI

XU

X

U

c

w

H

t s

(20)

Setting ,1== Rβ using (11) and simplifying yields, ( )φ−

∂∂

=∂

∂− 11

1 MI

pw

H

ts

(21)

From equation (19), (21) can be rewritten in the logarithmic form as,

( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( )φθθθα −−−−+++−++= 1lnln1lnlnlnlnlnlnln 12211 HMIwp M (22)

Recall that ,1=δ and thus, 1lnln HI = , so (22) can be rewritten as:

( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( )112111 ln1lnlnlnln1lnlnlnln HwpMH M −−−−−−−−+=− θθθαφ (23)

The reduced form of the demand for medical services can be obtained by the resolution of the following system of two equations:


11

( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( )

( ) ( ) ( ) ( ) ( ) ( )

−+−−+−−−−−=

−−−−−−−−+=−

M

MEMMM

M

EwpHM

HwpMH

α

αααφαα

θθθαφ

1lnln11ln1ln1lnln

ln1lnlnlnln1lnlnlnln

1

112111

(24)

Recall that the second equation of this system is the structural demand for medical services as in (17). The reduced form of the demand for medical services is given by:

( ) ( )( ) ( ) ( )( ) ( ) ( )( ) ( ) ( ) EwpM EMMM αεεαεαφεα −−−++−+−−−+−Θ= 1ln11ln111ln11ln (25)

Where ( ) 11/1 2 <+≡ θε and 10 www =≡ .

In contrast to the structural demand function for health in (17), this function does not depend on health stock. The positive relationship between remittances and the demand for medical services is reaffirmed. The

theoretical impact is given by ( )11 ++ εαM .

3.2. Empirical framework

A number of models are evaluated in an attempt to test the specified hypotheses. Hence, a base model would be modified to fit each one of the specific hypotheses.

The base model

Specification

The primarily concern is to determine the differential effects of remittances on access to public and private health care services. We depart from an econometric model that relates remittances to the uses of those services while controlling for other potential determinants of the uses of health care services. Based on some recent studies on the access to health care services in developing countries (Gwatkin et al., 2007; Berthelemy & Seban, 2009), the following model is specified:

sd

ijtijt

sd

jt

sdsd

ijt

sd

ijt rhfXhsa,,,,,

εµδφβ ++++′=

The variables are all expressed in logarithm term with, sd

ijthsa,

representing the percentage of children of

the ith quintile that had access to a health care service of the type s (public or private) in the treatment of illness d in the jth country in the tth year. X, hf and r represent the vector of control variables (education of mothers and per capita income), the other sources of financing health care (public spending and foreign

health aid) and finally remittances flows in each country, respectively. iµ is the quintile fixed effect and

ijtε is a white noise disturbance. In the model, we control for the education of the mother as it is a key

determinant of the child’s health. We control as well for the level of development given that it logically exists a relationship between a country’s degree of development and its populations’ access to health care services.

Two coefficients are of a special interest in the current specification: the coefficient associated with the

remittances ( )δ and the coefficient associated with the other sources of financing ( )φ .


12

A serious problem in these kinds of specifications is the endogeneity of remittances, foreign aid and of public spending, as pointed out by Chauvet and al. (2008).

Endogeneity of remittances, foreign aid and public spending in health

Sources of endogeneity

A naïve estimation of the parameters of r and of hf, may lead to biased estimates if the issue of endogeneity is not taken into account. Three sources of endogeneity are generally observed: a measurement error on the considered variables, a reverse causality bias and an omitted variables bias. As far as remittances are concerned, officially recorded values are known to be measured with error. Estimates of unrecorded remittances range from 20 to 200 percent of official statistics on remittances (Freund & Spatafora, 2008). Balance of payment statistics produced by developing countries often neglect remittances received via money transfer operators and almost always exclude those transferred via informal means such as hawala operators, friends, and family members.

Moreover, if remittances are targeted by the migrants to directly finance the health of their siblings, this would result in a reverse and negative causality between remittances and the access to health care services in the sense that remittances would respond to a low level of household’s access to health care services.

Regarding the omitted variable bias it could exist a third variable, which would simultaneously affect the level of remittances received by households and their level of access to health care services. That is especially the case for the intensity as well as of the quality of emigration in the country. For instance, international migration of physicians represents a migration of skilled individuals which in return has a clear effect on the amount of remittances received by the considered country (Adams, 2009). However, the severity of the medical brain drain may reduce the access of populations to health structures for their needs for treatment. International migration has another effect in the considered model: the effect on fiscal revenue. Indeed, although migration of individuals is positively correlated to the level of remittances, it reduces the number of tax payers which in turn leads to a lower amount of tax revenue and thus to a decrease in the level of public spending on social sectors.

The endogeneity of public spending as well as of foreign aid to the health sector may be due to a measurement error (which is more pronounced in the case of disaggregated levels of foreign aid) on the one hand, and to a simultaneity bias on the other hand. In effect, both public spending and foreign health aid are targeted to improve population access to medical services and to ensure better quality of treatments provided. In an econometric perspective, this positive reverse causality between the access to health care services and the sources of funding leads to severe estimations problems in the parameters.

Factoring in the endogeneity of remittances, public spending on health and health aid

The endogeneity of the considered variables can be dealt with in various ways. A first approach is that of a selection on observables. Such an approach consists in adding a large number of control variables. Such variables are supposed to be strongly correlated with the variables of interest as well as with the dependent variable. Although this approach is quite easy to implement, the strategy is based on the assumption that the researcher has controlled for all the variables such that the hypothesis of the orthogonality of the endogenous variables of interest with the error term is not rejected. However, it is impossible to think that with this approach one can control for all the implied bias. Moreover, the choice of the control variables is often conditioned by the availability of data on the control variables.

A second approach is based on the selection on unobservables instead. This strategy consists in the introduction in the model of fixed effects which in fact measure all other (time-invariant) determinants of access to health care services and for which a direct measure is not possible. In the current paper, we purport


13

to introduce quintile fixed effects to control for the heterogeneity between income quintiles. However, we cannot introduce country fixed-effect given that about 60% of our database is made up of countries with at most one year of observations on the dependent variables.

A third approach for treating for the endogeneity of the included variables is to resort to instrumental variables. A valid instrument for the variables of interest would have to satisfy three conditions. First, it would need to be correlated with the supposed endogenous variable. Second, the measurement error in the instruments would need to be uncorrelated with the measurement error in endogenous variables. Third, the instrument should not be correlated with the access to health care services, except through the endogenous variables or through the effect on the other variables that have been already controlled for. The choice of pertinent instruments for each one of the endogenous variables of our interest (remittances, public spending and foreign aid) is driven by the recent literature. As in Chauvet and al. (2008), remittances would be instrumented by the level of financial development (the ratio of broad money supply (M2) to GDP). Indeed, the greater the level of financial development, the higher the amount of remittances received, given the reduction of the transaction costs induced by the financial widening (Freund & Spatafora, 2008).

Following Filmer & Prichett (1999) and Bokhari and al. (2007), we use as instruments for public spending on health, the per capita public spending on the health sector, and the per capita spending on defense of a country's geographic neighbors. The higher is spending on defense in neighboring countries, the higher the share of military spending in the considered country’s budget and thus, the lower the level of public spending on social sectors. However, it tends to be a positive correlation between spending on health in neighboring countries and spending on health in the considered country.

Concerning foreign aid on health sector, we follow the instrumentation procedure initiated by Tavarez (2003) and recently revisited by Brun and al. (2006) and Chauvet and al. (2008). The idea behind this procedure is that the level of foreign aid received by a given country from each one of the major donors is related to the various aspects of the proximity (geographical and cultural proximities) between the considered developing country and the donors. We retain 21 major OECD countries as principal sources of foreign aid to developing countries. To address the issue of geographical proximity, we use the inverse distance between a country and each of the 21 donors. Total aid send by each donor is then, weighted by the inverse geographical distance. To take into account cultural proximity, we use a dummy variable which takes 1 if the country shares the same language with the donor. We also use a dummy variable for the religious proximity between the donor and the recipient. The common language or common dominant religion dummies are used as weights for the amount of aid given by each donor. Altogether, foreign aid received by each country is instrumented by three variables.

The quality of the considered instruments should be confirmed by the performance of the specified models. On that respect, we would rely on the Stock and al. (2002) weak-instruments test and the standard over identification tests to empirically justify their use.

Are remittances complements or substitutes of foreign health aid? An econometric model of the

interaction between external sources of funding

In order to test the existence of a complementary or substitutability relationships between external sources of funding in developing countries (remittances and foreign health aid), we modify the base model by adding an interaction term of remittances and health aid per capita. The following model is specified:

( ) d

ijtijtjt

d

jt

d

jt

dd

ijt

d

ijt hfrrhfXhsa εµδδφβ +++++′= *21

Two distinct results may derive from the estimation of this model. On the one hand, there could exist a complementary effect between remittances and health aid. In this case, the following condition should hold:


14

,0, >sdφ 0,

1 >sd

δ and .0,

2 >sd

δ On the other hand, there could be a substitution effect. In that case, we

should observe that ,0, >sdφ 0,

1 >sd

δ and .0,

2 <sd

δ

However, we must be cautious when interpreting the results of the model. Indeed, a low level of health aid per capita might measure either an increase of the financial constraints for a recipient country or an increase in the financial autonomy of the country. Thus, we have to take into account this duality in the econometric model. To achieve this objective, we make the assumption that a diminution in the amount of health aid in a low income country reflects the tightening of financial constraints faced by this poor country whereas it not necessarily the same interpretation for an emerging country. From an economic sense, we are testing the hypothesis that the nature of the interactions between remittances and foreign aid might be conditioned by the level of economic development (income per capita). Our preferred econometric model is then a model which allows the inclusion of a double interaction between remittances, health aid and income per capita. The following specification holds:

( ) ( ) ( )jtjt

sd

jtjtjt

sd

jtjt

sd

jt

sd

jt

sdsd

ijt

sd

ijt yryAidrAidrrAidXhsa ****,

4

,

3

,

2

,

1

,,,δδδδφβ +++++′=

( ) sd

ijtijtjt

sdyAid

,,

5 * εµδ +++

We are particularly interested by the cross derivative of the access to health care services with respect to

remittances and foreign health aid

+=

∂∂

∂jt

dd

jtjt

sd

ijty

Aidr

hsa32

,2

δδ .

This calculus is a way to measure empirically the substitution or the complementary effect between

remittances and aid given different levels of income per capita. When d

2δ and d

3δ have opposite signs

( 02 >d

δ and )03 <d

δ , a threshold effect of income per capita expressed in per capita dollars terms

=

−d

d

ey 3

2

* δ

δ

arises. Under these conditions, we can therefore conclude that remittances are a complement

of health aid in poor countries and this complementary effect disappears progressively when the level of per capita income rises.

The duality behind the nature of the interactions between remittances and health depending on the level of per capita income is indeed straightforward. The hypothesis that remittances and foreign aid are complements in poor countries and substitutes in relative rich countries simply reflects the idea that poor countries need both remittances and health aid to achieve high rates of utilization of modern health care services while relative rich ones do not need absolutely the two flows.

Heterogeneity in the remittances’ impacts on the access to health services by income quintiles

If median and richer classes are the primarily and the most concerned by international migration (given the cost to migrate abroad), then remittances in return are received in a larger amount by these classes in a considered country. One would expect that the impact of remittances on the access to health will be the


15

greatest for the third, the fourth and the fifth income quintiles. If this result is confirmed, it will posit remittances as an imperfect substitute for the other types of financing health sector in developing countries. Moreover, if remittances induce a sectorial glide in the utilization of health structures from the public to the private sector, this result should hold for the median and richer quintile in the distribution of income. As households receive remittances they would increase their search for better quality in health care delivery. Such income effects induced by remittances on the behavior of receiving households need to be taken into account in developing country. Then, if remittances do not crowd-out public funds in health sector, one may expect that public subsidies in the health sector will be received by the poorest of the remittance receiving countries.

In order to test the hypothesis of heterogeneity of the impact of remittances on the access to health care services over the income quintiles, the following specification is adopted:

( ) sd

ijti

i

ijt

sd

ijt

sd

jt

sdsd

ijt

sd

ijt rrhfXhsa,

5

2

,,,,,* εµµϕδφβ +++++′= ∑

=

where iµ is the quintile dummy variable, i is the income quintile index. We expect that all the 5,4,3=iϕ

parameters will be negative in the regression for the public sector and positive in the regression for the private sector.

4. DATA

Variables on the access to health care services

Data on variables on the access to health care services are taken from the study by Gwatkin and al. (2007) on Health, Nutrition and Population in 56 developing countries, and all the data are disaggregated by income quintiles2. In this database, more than half of the countries are African. The report of Gwatkin et al. (2007) is based on data drawn from several demographic and health surveys (DHS) conducted in these countries. These surveys target especially maternal and child health with a standardized questionnaire. Data also include socioeconomic variables like mother education for each quintile.

The report includes several indicators of health status and utilization of health services. For our work, we are only interested in the second type of indicators (access to health). Two variables on the access to health care services are chosen for our purpose:

� Treatment of fever: percent of children with fever, with or without cough or rapid breathing, in the two weeks before the survey who had sought medical advice for fever, at any medical facility or provider, whether public or private,

� Treatment of diarrhea : percent of children with diarrhea in the two weeks before the survey who had been taken for treatment, at any medical facility or provider, whether public or private,

In our analysis, we also differentiate the treatment of each disease in a public structure from that in a private structure. The geographical coverage of these data is defined in the appendix Table 1.

Per capita public spending on health sector

Data on public spending are drawn from several sources: World Development Indicators, OECD Health Database, WHO World Health Statistics, OECD Health Database and IMF Government Financial Statistics.

2 Other studies have used this database in order to analyze the determinants of child-health outcomes and access to health in developing countries (Fay and al., 2005; Ravallion, 2007; Berthelemy & Seban, 2009).


16

Public spending in health for each country are in $US and divided by the total population of each country. Data on population comes from World Development Indicators.

Development assistance on health (health aid)

The existing research on global health resource flows has yielded some important estimates and findings, but it does not provide comprehensive and systematic estimates of development assistance on health (DAH) over an extended period of time. A majority of studies have relied on databases maintained by the Development Assistance Committee of the Organization for Economic Cooperation and Development (OECD-DAC). While these databases are a valuable source of information, they do not capture all the external aid on health sector. The biggest gap in coverage stems from the fact that the databases only reflect official development assistance (ODA) flowing from governments and leave out key private actors in the health domain like the Bill & Melinda Gates Foundation (BMGF), other private foundations, and non-governmental organizations (NGOs).

The Institute for Health Metrics and Evaluation (IHME) has launched a multi-year program for tracking DAH, which has addressed these conceptual and measurement challenges and developed a comprehensive system for global health resource tracking. The primary goal of the program is to develop consistent time-series data on DAH, which will be updated annually. The beginning year in this database is 1990. Development assistance for health is defined as financial and in-kind contributions made by channels of development assistance to improve health in developing countries. It includes all disease-specific contributions as well as general health sector support, and excludes support for allied sectors. Data are in US dollars per capita terms.

International migrants’ remittances Remittances data are drawn from the World Bank World Development Indicators Tables. Remittances are defined as the sum of three components: workers’ remittances, compensation of employees and migrants’ transfers. Workers’ remittances are classified as current private transfers from migrant workers who are residents of the host country to recipients in their country of origin. They include only transfers made by workers who have been living in the host country for more than a year, irrespective of their immigration status. Compensation of employees is the income of migrants who have lived in the host country for less than a year. Migrants’ transfers are defined as the net worth of migrants who are expected to remain in the host country for more than one year that is transferred from one country to another at the time of migration. This is the most reliable database on remittances for cross-country analysis. Data in the study are in US dollars per capita terms. Military expenditures

Data are drawn from the Stockholm International Peace Research Institute (SIPRI). This institution publishes data on military spending for 172 countries since 1988. We use military spending in US dollars per capita terms. Income per capita GDP per capita in US dollars constant prices is used as a proxy for the level of development. The data come from the World Bank World Development Indicators. Table of descriptive statistics of variables is presented in Table A1 in Appendix.

5. ECONOMETRIC RESULTS

Now, we turn to the estimation and the discussion of econometric results.


17

Impact of remittances, health aid and public spending on access to health care services This subsection is devoted to the comparison of the respective impacts of remittances, health aid and public spending on the access to health services in developing countries. Tables 3 and 4 summarize these results. Table 3 presents the results of the impact of remittances and health aid on access to health services for each disease (fever and diarrhea). For each disease, we use as dependent variable, the total access to health services, the access to public and the access to private services. We are also interested in the validity of our instrumental variable strategy. Two tests are performed in order to check the quality of our instruments. As we can see in Table 3, the F-test for the first-stage regression and the Hansen suridentification test (OID) confirm the good quality of our instrumentation3. Regarding the impact of our variables of interest, we find that the elasticity of access to health services with respect to remittances is positive and statistically significant for the total and private health services access. In contrast, remittances appear to have a non significant effect on the access to public health care services (column 2) and a negative effect in column (5).

Table 3: Effects of Remittances and Health aid on access to health care services

Dependent variables : Access to health services by diseases and sectors

Fever

Fever_pub

Fever_pri

Diarrhea

Diarr_pub

Diarr_pri

(1) (2) (3) (4) (5) (6)

Remittances per capita (log) 0.313*** 0.0191 0.781*** 0.152** -0.153** 0.661*** (2.704) (0.242) (3.550) (2.226) (-2.270) (4.379) Health aid per capita (log) 1.159*** 0.997*** 0.954 0.105 0.0147 -0.0356 (2.909) (3.580) (1.300) (0.546) (0.0791) (-0.0782) GDP per capita (log) -0.935 -0.719 -0.690 -0.104 -0.0722 -0.739 (-0.837) (-0.860) (-0.330) (-0.152) (-0.109) (-0.474) (log(GDP per capita))² 0.0543 0.0625 -0.000510 -0.00222 0.0163 0.00414 (0.688) (1.058) (-0.00346) (-0.0459) (0.353) (0.0372) Mother education 0.310*** 0.213*** 0.432*** 0.272*** 0.207*** 0.491*** (5.632) (4.607) (4.845) (7.009) (5.017) (6.348) Quintile_2 0.0301 0.0310 0.0636 0.000712 -0.0290 0.0617 (0.238) (0.328) (0.286) (0.00973) (-0.346) (0.379) Quintile_3 0.0459 0.0470 0.134 0.0120 -0.0426 0.0427 (0.354) (0.482) (0.603) (0.153) (-0.505) (0.242) Quintile_4 0.0799 0.00910 0.245 -0.00198 -0.130 0.107 (0.580) (0.0863) (1.041) (-0.0244) (-1.506) (0.606) Quintile_5 0.163 -0.123 0.622** 0.0808 -0.247** 0.432** (1.061) (-1.026) (2.365) (0.890) (-2.386) (2.140) Constant 4.034 3.174 1.761 2.716 2.492 3.057 (1.192) (1.229) (0.281) (1.341) (1.246) (0.666)

Observations 380 380 378 392 392 383 Remittances instrumentation F-stat 34.80 34.80 33.94 37.40 34.80 39.18 Health aid instrumentation F-stat 16.07 16.07 15.49 18.93 16.07 18.66 Hansen OID p-value 0.86 0.11 0.13 0.34 0.07 0.94

Note : All variables in the models are in log terms except income quintiles dummies. Fever_pub, Fever_priv, Diarrhea_pub, Diarrhea_priv express the access to health services for the treatment of fever and diarrhea in public and private facilities. Mother education is available by quintiles. Remittances are instrumented by the level of financial development (ratio of broad money M2 to GDP) and health aid is instrumented by the total aid budget of the 21 main donors weighted by dummies variables which take the value 1 if donor and recipient countries share the same language or the same religion and 0 otherwise. Robust t-statistics are in parentheses. Standard errors clustered at country level. * significant at 10%; ** significant at 5%; *** significant at 1%.

Furthermore, health aid contributes to increase the total level of access to health as well as the access to public health services. These results are in conformity with the hypothesis formulated previously. Indeed, remittances improve overall access to health services in developing countries, but also induce a sectorial

3 The instrumentation equations are presented in Tables A1-2 in Appendix.


18

glide of frequentation from the public sector to the private sector. We can interpret this result with the hypothesis of a search by recipient households, for higher quality of health services when their income increases. In contrast, the sectorial impact of health aid is in favour of the public services access because aid is mainly devoted to this sector. We can also notice that the effect of health aid on the overall access to health care services is larger than the impact of remittances. This is not a surprising result given that remittances are not received by the whole population in a country while foreign aid might theoretically be consumed by all fringes of the population. Interesting results are also found regarding control variables. For example, mother education is a major determinant of child access to health care services. Table 4 presents the results of estimations when health aid is replaced by public spending in the model. We cannot include simultaneously the two variables because of colinearity. Indeed, health aid is financing a non negligible part of public spending in developing countries. Results are quasi similar to those of the previous model. Remittances induce a sectorial glide in the frequentation from public to private sector while public spending increase overall access to health services as well as access to public services. Moreover, the impact of remittances although large and significant on the overall access is lower than that of public spending on health. Following the results of diagnostic tests performed, our instrumentation is valid only for diarrhea equations.

Table 4: Effects of Remittances and Public Health Spending on access to health care services

Dependent variables : Access to health services by diseases and sectors

Fever

Fever_pub

Fever_pri

Diarrhea

Diarrhea_pub

Diarrhea_pri

(1) (2) (3) (4) (5) (6)

Remittances per capita (log) 0.217** 0.0103 0.502*** 0.275*** 0.0163 0.604*** (2.325) (0.103) (3.440) (2.844) (0.153) (4.299) Public health spending per capita (log) 0.365*** 0.745*** -0.982*** 0.380*** 0.882*** -0.834*** (3.861) (5.690) (-5.513) (3.018) (6.159) (-4.337) GDP per capita (log) 2.711*** 2.178** 1.954 0.636 0.204 -0.504 (3.338) (2.214) (1.496) (0.903) (0.274) (-0.377) (log(GDP per capita))² -0.254*** -0.225*** -0.0945 -0.101* -0.0961* 0.0713 (-4.096) (-3.037) (-0.961) (-1.809) (-1.779) (0.687) Mother education 0.220*** 0.104* 0.390*** 0.270*** 0.214*** 0.523*** (4.168) (1.791) (3.751) (5.518) (3.435) (5.352) Quintiles_2 0.0685 0.0732 0.0612 0.00294 -0.0306 -0.000208 (0.646) (0.640) (0.324) (0.0271) (-0.262) (-0.00119) Quintiles_3 0.107 0.127 0.121 0.0518 -0.0299 0.0753 (0.997) (1.115) (0.621) (0.465) (-0.243) (0.407) Quintiles_4 0.155 0.0960 0.320 0.00144 -0.154 0.0453 (1.340) (0.767) (1.520) (0.0115) (-1.158) (0.228) Quintiles_5 0.330** 0.0934 0.669*** 0.134 -0.211 0.402* (2.535) (0.619) (2.712) (0.998) (-1.445) (1.790) Constant -5.573** -3.840 -6.835* 0.971 2.728 0.949 (-2.172) (-1.225) (-1.658) (0.445) (1.148) (0.228)

Observations 218 218 218 225 225 223 Remittances instrumentation F-stat 10.88 10.88 10.88 11.47 11.47 12.73 Public spending instrumentation F-stat 52.60 52.60 52.60 51.02 51.02 50.19 Hansen OID p-value 0.001 0.003 0.028 0.18 0.097 0.244

Note : All variables in the models are in log terms except income quintiles dummies. Fever_pub, Fever_priv, Diarrhea_pub, Diarrhea_priv express the access to health services for the treatment of fever and diarrhea in public and private facilities. Mother education is available by quintiles. Remittances are instrumented by the level of financial development (ratio of broad money M2 to GDP) and public sector health spending is instrumented by the average public sector health spending per capita, and the average

defense spending per capita of a country’s geographic neighbors. Robust t-statistics are in parentheses. Standard errors clustered at country level. * significant at 10%; ** significant at 5%; *** significant at 1%.


19

Are external sources of funding complements or substitutes? Empirical results of the interactions between

remittances and health aid in developing countries

In this section, we test the hypothesis that the interactions between remittances and foreign health aid might vary with the level of economic development in recipients’ countries. More precisely, we test that remittances and foreign health aid are complements in relatively poor countries and somewhat substitutes in relatively rich ones.

Two econometric estimations are performed. Table 5 presents the results in the case of the two diseases retained: fever and diarrhea. The dependent variable is the overall access to health care services and our variables of interest are instrumented adequately. While interesting and similar results are observed in the both two columns, we restrict the interpretation of the results to those of column 1 in which all the diagnostic tests performed are conclusive (in column 2, the low probability taken by the Hansen OID test suggests that in this specification, our instruments are not valid).

When we look at the main results of the estimation in column 1, we notice that the hypothesis formulated here is confirmed empirically. Indeed, the cross derivative of the access to health care services (for the treatment of fever) with respect to remittances and health aid (4.22 – 0.63*log (Health aid)) is statistically significant and is conditional to the level of economic development. An optimal threshold value of per capita income arises and is evaluated at around 812$US per capita per year. Putting it differently, the cross derivative of the access to health care services is positive at low levels of income, precisely when the level of per capita income is less than 812$US. Above this value, the derivative turns to be negative.

These results suggest that remittances and foreign health aid appear to be complements in low income countries and substitutes in the others. This conclusion holds because one can think that low income countries need multiple sources of funding for their development given the high levels of needs they are faced with. Remittances and health aid when combined are thus crucial to reduce the levels of financial constraints and then, to achieve high rates of access to medical services whatever the sector. In contrast, in relatively “rich” countries, remittances and foreign health aid act as substitutes. In these countries, the achievement of high rates of utilization of medical services is not absolutely conditioned by the inflows of both remittances and health aid.

Then, even if remittances may constitute an important and stable resource for development in low income countries, health aid doesn’t be reduced in response. In such countries, a special combination of remittances and foreign aid is most than ever appropriate to increase the accumulation of human capital. This is an important conclusion in the actual international debate on the financing of development.

Heterogeneity in the remittances’ impact on access to health services by income quintiles: Econometric

results

We have previously found that remittances lead to a sectorial glide in the uses of health care services from the public to the private sector (see Table 3 and 4). However, it is interesting to explore whether this shift comes from the behaviour of a particular income class. More precisely, we test the hypothesis that the sectorial glide in frequentation of health care services is due to the intermediate and rich classes who logically benefit the most from remittances.

Table 6 summarizes the results of estimations. The first four columns consist in the estimation of the econometric model which includes as control, health aid. When we look at the first two columns (the fever case), we observe that the coefficient of the interactions between remittances and the quintiles 3, 4 and 5 are statistically significant and negative in the public sector equation and positive in the case of private sector model.


20

These results are also validated in the case of diarrhea but only for quintiles 4 and 5. We can then conclude that remittances recipients’ households modify their behaviour by leaving the utilization of public services to experiment the private one from which they expect a better quality of services. The last four columns present the results of the estimations of models in which we control for the impact of public spending. Similar tendencies are obtained. Remittances continue to lead to the sectorial shift in frequentation even if in this specification, the effect appears to concern only the richest class.

Table 5: Remittances and Health Aid: Complements or Substitutes? Dependent variables : Access to health services by diseases (1) (2) Fever Diarrhea

Remittances per capita (log) -3.699** -2.907** (2.38) (2.46) log(Remittances per capita)*log(Health aid per capita) 4.219*** 2.306* (2.90) (1.93) log(Remittances per capita)*log(Health aid per capita)*log(GDP per capita) -0.630*** -0.395** (3.11) (2.38) log(Remittances per capita)*log(GDP per capita) 0.580** 0.488*** (2.54) (2.79) log(Health aid per capita)*log(GDP per capita) 1.423*** 1.293*** (2.87) (2.81) Health Aid per capita (log) -8.652*** -6.656** (2.93) (2.45) GDP per capita (log) -2.459* -0.001 (1.92) (0.00) (log(GDP per capita))² 0.074 -0.112 (0.75) (1.02) Mother education 0.387*** 0.206*** (6.81) (3.79) Quintile_2 0.010 0.012 (0.10) (0.11) Quintile_3 -0.004 0.047 (0.04) (0.42) Quintile_4 -0.002 0.060 (0.02) (0.50) Quintile_5 0.058 0.158 (0.45) (1.15) Constant 13.798*** 5.599 (2.88) (1.18)

Observations 380 392 Joint significance probability of remittances’ coefficients 0.000 0.015 Threshold level of income per capita ($US constant term) 812 341


21

Remittances instrumentation equation F-stat 31.764 30.854 Remittances*Health aid instrumentation equation F-stat 14.150 14.209 Remittances*Health Aid*GDP per capita instrumentation equation F-stat 16.907 16.183 Remittances*GDP per capita instrumentation equation F-stat 32.743 31.704 Health Aid*GDP per capita instrumentation equation F-stat 15.338 15.879 Health Aid instrumentation equation F-stat 14.770 14.822 Hansen OID p-value 0.100 0.002 Note : All variables in the models are in log terms except income quintiles dummies. Fever and Diarrhea express the access to health services for the treatment of fever and diarrhea in modern facilities. Mother education is available by quintiles. Remittances are instrumented by the level of financial development (ratio of broad money M2 to GDP) and health aid is instrumented by the total aid budget of the 21 main donors weighted by dummies variables which take the value 1 if donor and recipient countries share the same language or the same religion and 0 otherwise. Interactive terms are instrumented by the product of the respective instruments of each variable. Robust t-statistics are in parentheses. Standard errors clustered at country level. * significant at 10%; ** significant at 5%; *** significant at 1%.

For robustness checks, we add also to the previous specification the interaction terms of health aid per capita and each of the four quintiles dummies. Two important advantages arise from this specification. Firstly, it allows us to test the robustness of the previous results in which only remittances was interacted with quintiles dummies. Secondly, we can examine whether some impacts of health aid might also vary among quintiles of income. Results are presented in Table 7. The principal message is that the distributional and sectorial impacts of remittances remain unchanged whereas no pro-poor and no pro-rich effects of health aid are observed from our data.

Table 6: Heterogeneity of remittances’ impact on access to health services among income quintiles Dependent variables : Access to health services by diseases and sectors

Fever_ pub

Fever_ priv

Diarrhea_pub

Diarrhea_ priv

Fever_ pub

Fever_ priv

Diarrhea_pub

Diarrhea_ priv

(1) (2) (3) (4) (5) (6) (7) (8)

Remittances per capita (log) 0.128 0.323 -0.110 0.230 0.142 0.325 0.158 0.409 (0.721) (1.412) (-1.147) (1.092) (0.632) (1.244) (0.626) (1.617) Remittances*Quintile_2 0.0123 0.0565 0.00367 0.0167 0.00620 0.102 -0.0381 0.126 (0.241) (0.963) (0.0612) (0.207) (0.104) (1.140) (-0.503) (0.954) Remittances*Quintile_3 -0.0852* 0.0938 -0.0659 0.00888 -0.0844 0.122 -0.135 0.0252 (-1.883) (1.091) (-1.020) (0.0934) (-1.343) (0.895) (-1.253) (0.260) Remittances*Quintile_4 -0.219*** 0.318** -0.152* 0.307** -0.151* 0.258 -0.157 0.403** (-2.621) (2.463) (-1.873) (2.574) (-1.818) (1.456) (-1.343) (2.241) Remittances*Quintile_5 -0.399*** 0.371** -0.376*** 0.307* -0.494*** 0.455* -0.428*** 0.489** (-4.007) (2.444) (-3.524) (1.958) (-3.216) (1.847) (-3.016) (2.145) Health aid per capita (log) 1.277 -0.361 0.253 -0.740 (1.143) (-0.327) (0.412) (-0.713) Public health spending per capita (log) 0.723** -0.965*** 0.867*** -0.820** (2.164) (-2.825) (2.627) (-2.478) GDP per capita (log) -0.520 1.464 0.311 1.584 2.408 1.782 0.322 -0.647 (-0.244) (0.587) (0.253) (0.664) (1.252) (0.702) (0.253) (-0.245) (GDP per capita (log))² 0.0515 -0.146 -0.00465 -0.152 -0.237 -0.0851 -0.101 0.0781 (0.334) (-0.818) (-0.0512) (-0.876) (-1.583) (-0.441) (-1.034) (0.380) Mother education 0.169* 0.438*** 0.151** 0.436*** 0.0610 0.422** 0.178 0.564*** (1.647) (3.542) (1.998) (3.606) (0.553) (2.178) (1.612) (3.112) Quintile_2 0.00975 -0.0650 -0.0209 0.0408 0.0718 -0.191 0.0745 -0.317 (0.0741) (-0.397) (-0.130) (0.179) (0.470) (-0.807) (0.373) (-0.909) Quintile_3 0.285*** -0.102 0.156 0.0592 0.357** -0.191 0.318 -0.0116 (2.666) (-0.430) (0.954) (0.237) (2.218) (-0.493) (1.279) (-0.0427) Quintile_4 0.597*** -0.548 0.298 -0.598* 0.498** -0.326 0.256 -0.953* (2.771) (-1.397) (1.347) (-1.714) (2.566) (-0.603) (0.804) (-1.920) Quintile_5 0.929*** -0.297 0.749*** -0.248 1.336*** -0.463 0.857*** -0.818 (3.557) (-0.599) (2.647) (-0.493) (3.595) (-0.604) (2.585) (-1.151) Constant 1.922 -3.398 0.715 -3.300 -4.913 -5.843 1.998 1.891 (0.301) (-0.447) (0.201) (-0.474) (-0.808) (-0.735) (0.498) (0.228)


22

Observations 380 378 392 383 218 218 225 223 Joint Signi. of Remittances coeff. p-value

0.002 0.025 0.000 0.007 0.018 0.330 0.009 0.215

Remittances instrumentation F-test 10.83 9.74 9.29 8.82 2.95 2.95 2.48 2.33 Remittances*Quintile_2 instrum. F-test 7.43 7.05 8.28 9.29 1.99 1.99 2.12 2.46 Remittances*Quintile_3 instrum. F-test 7.61 7.09 9.79 9.50 2.02 2.02 2.42 2.47 Remittances*Quintile_4 instrum. F-test 8.31 7.88 9.11 10.58 2.04 2.04 2.47 2.52 Remittances*Quintile_5 instrum. F-test 10.17 13.18 13.15 15.33 2.69 2.69 2.98 3.14 Health aid instrumentation F-test 2.73 0.92 1.37 0.78 .. .. .. .. Public health spending instrum. F-test .. .. .. .. 5.15 5.15 5.18 5.06 Hansen OID p-value 0.64 0.23 0.14 0.05 0.13 0.29 0.34 0.57

Note: All variables in the models are in log terms except income quintiles dummies. Mother education is available by quintiles. Remittances are instrumented by the level of financial development (ratio of broad money M2 to GDP); public sector health spending is instrumented by the average public sector health spending per capita, and the average defense spending per capita of a country’s geographic neighbors; health aid is instrumented by the total aid budget of the 21 main donors weighted by dummies variables which take the value 1 if donor and recipient countries share the same language or the same religion and 0 otherwise. Interaction terms are instrumented by the product of instruments of remittances with quintiles dummies. Robust t-statistics are in parentheses. Standard errors are clustered at country level. * significant at 10%; ** significant at 5%; *** significant at 1%.


23

Table 7: Heterogeneity of remittances’ and Health aid impacts on access to health services among income quintiles Dependent variables : Access to health (1) (2) (3) (4) services by diseases and sectors Fever_pub Fever_pri Dia_pub Diar_pri

Remittances per capita (log) 0.163 0.276 -0.088 0.186 (1.06) (1.55) (0.80) (1.10) Remittances (log)*Quintile_2 0.012 0.040 -0.015 -0.036 (0.06) (0.21) (0.12) (0.17) Remittances (log)*Quintile_3 -0.068 0.135 -0.069 0.015 (0.37) (0.76) (0.58) (0.08) Remittances (log)*Quintile_4 -0.228 0.438** -0.189* 0.375** (1.49) (2.05) (1.67) (1.99) Remittances (log)*Quintile_5 -0.405*** 0.490** -0.402*** 0.375* (2.75) (2.07) (2.61) (1.73) Health aid per capita (log) 1.830* -1.613 1.243 -1.935 (1.79) (1.26) (1.60) (1.61) Health aid per capita (log)*Quintile_2 0.010 0.250 -0.498 -0.105 (0.01) (0.16) (0.50) (0.06) Health aid per capita (log)*Quintile_3 -0.220 1.119 -0.553 0.866 (0.16) (0.72) (0.58) (0.55) Health aid per capita (log)*Quintile_4 -0.849 2.384 -1.040 1.858 (0.70) (1.44) (1.12) (1.30) Health aid per capita (log)*Quintile_5 -1.365 2.730 -2.102* 2.479 (1.16) (1.63) (1.85) (1.63) GDP per capita (log) -0.883 1.417 0.294 1.987 (0.87) (1.07) (0.40) (1.62) (GDP per capita (log))² 0.078 -0.145 -0.002 -0.184** (1.07) (1.54) (0.04) (2.07) Mother education 0.154*** 0.495*** 0.127*** 0.469*** (2.71) (6.55) (2.83) (6.56) Quintile_2 0.001 -0.313 0.560 0.298 (0.00) (0.15) (0.43) (0.14) Quintile_3 0.493 -1.484 0.769 -0.912 (0.26) (0.74) (0.62) (0.46) Quintile_4 1.561 -3.559* 1.503 -2.860 (0.98) (1.66) (1.27) (1.57) Quintile_5 2.454 -3.678 3.110** -3.188 (1.56) (1.63) (2.12) (1.59) Constant 2.495 -1.773 -0.324 -3.225 (0.77) (0.42) (0.14) (0.84)

Observations 380 378 392 383 Joint Significance prob. of Remittances coefficients 0.005 0.000 0.000 0.001 Joint Significance prob. of Health aid coefficients 0.124 0.414 0.254 0.227

Remittances instrumentation F-stat 14.297 14.069 16.240 16.810 Remittances*Quintile_2 instrumentation F-stat 6.625 6.617 7.076 7.435 Remittances*Quintile_3 instrumentation F-stat 6.702 6.696 7.706 7.365 Remittances*Quintile_4 instrumentation F-stat 6.573 6.700 7.317 7.213 Remittances*Quintile_5 instrumentation F-stat 6.660 6.660 7.537 7.267 Health aid instrumentation F-stat 1.811 1.931 1.896 1.959 Health aid*Quintile_2 instrumentation F-stat 0.610 0.612 0.610 0.550 Health aid*Quintile_3 instrumentation F-stat 0.476 0.476 0.521 0.524 Health aid*Quintile_4 instrumentation F-stat 0.324 0.465 0.405 0.529 Health aid*Quintile_5 instrumentation F-stat 0.522 0.522 0.524 0.523 Hansen OID p-value 0.770 0.579 0.135 0.050 Note: All variables in the models are in log terms except income quintiles dummies. Mother education is available by quintiles. Remittances are instrumented by the level of financial development (ratio of broad money M2 to GDP); health aid is instrumented by the total aid budget of the 21 main donors weighted by dummies variables which take the value 1 if donor and recipient countries share the same language or the same religion and 0 otherwise. Interaction terms are instrumented by the product of instruments of remittances and health aid with quintiles dummies. Robust t-statistics are in parentheses. Standard errors are clustered at country level. * significant at 10%; ** significant at 5%; *** significant at 1%.


24

6. CONCLUDING REMARKS

This paper has examined the impacts of remittances, health aid and public health spending on access to health services in developing countries using recent intra-country data and instrumental variables methodology.

We have investigated the respective impacts of these variables on the access to health. Globally, we find that remittances, health aid and public spending are important determinants of the access to health services in recipients’ countries. Another interesting result comes from the fact that, remittances lead to a sectorial glide in the uses of health care services from the public to the private sector for the intermediate and richest income class that are the major recipients of remittances in developing countries. Moreover, remittances and foreign health aid are complements for the access to health care services in “low” income countries.

In terms of policy implications, several important conclusions can be formulated. Firstly, all strategies designed to increase the reception of remittances by households are welcome because these funds help households in poorer countries to use formal health services. Secondly, the “optimal therapy” for the low income countries is the combination of remittances and foreign aid. Then, governments and donors should not reduce their spending on the health sector in low income remittances’ recipient countries. Thirdly, considering the weak rates of utilization of health care facilities in developing countries in general and in African countries in particular, the role of remittances is also important given that, even if they certainly contribute to increase the access of the intermediate and the richest classes of income to private health care services, they make room for the poorest to access the public health care facilities. Therefore, remittances contribute globally to the improvement of the health access of all the layers of the population.

References

Adams, R. & Page, J. 2005. Do international migration and remittances reduce poverty in developing countries? World Development, 33(10), 1645–1669. Adams Jr., R. H. 2009. The Determinants of International Remittances in Developing Countries. World Development, 37, 93-103 Amuedo-Dorantes, C., Sainz, T., & Pozo, S. 2007. Remittances and healthcare expenditure patterns of populations in origin communities: evidence from Mexico. Intergration & Trade, 27, 159–184. Bansak, C. & Chezum, B. 2009. How Do Remittances Affect Human Capital Formation of School-Age Boys and Girls? American Economic Review, 99(2), 145–148. Bhargava, A. & Docquier, F. 2008. HIV Pandemic, Medical Brain Drain, and Economic Development in Sub-Saharan Africa. The World Bank Economic Review, 22(2), 345. Berthelemy, J.-C. & Seban, J. 2009. Dépenses de santé et équité dans l’accès aux services de santé dans les pays en développement. Revue d’économie du développement, 23(1), 33–71.

Bloom, D. E., Canning, D. & Sevilla, J. 2004. The Effect of Health on Economic Growth: A Production Function Approach, World Development vol. 32, no. 1, pp. 1-13.


25

Bokhari, F., Gai, Y., & Gottret, P., 2007. Government health expenditures and health outcomes. Health Economics, 16(3). Borraz, F. 2005. Assessing the impact of remittances on schooling: The Mexican experience. Global

Economy Journal, 5(1), 1–30. Brockerhoff, M. 1990. Rural-to-urban migration and child survival in Senegal. Demography, (pp. 601–616). Brun, J., Chambas, G., & Guerineau, S. 2006. Aide et mobilisation fiscale dans les pays en développement. AFD Programme RCH Aide et politique budgétaire, (pp. 065–2006).

Bryant, J., Browne, A. J., Barton, C., Zumbo, B. D., 2002, Access to Health Care: Social Determinants of Preventive Cancer Screening Use in Northern British Columbia, Social Indicators

Research, Vol. 60, No. 1/3, pp. 243-262

Calero, C., Bedi, A., & Sparrow, R. (2008). Remittances, liquidity constraints and human capital investments in Ecuador. World Development.

Carstensen, K., & Gundlach, E. 2006. The Primacy of Institutions Reconsidered: Direct Income Effects of Malaria Prevalence, The World Bank Economic Review, vol. 20, N° 3.

Castro-Leal, F., Dayton, J., Demery, L., & Mehra, K. 1999. Public social spending in Africa: Do the poor benefit? World Bank Research Observer, 14(1), 49–72. Chauvet, L., Gubert, F., & Mesple-Somps, S. 2008. Are remittances more effective than aid to improve child health. In ABCDE Conference (pp. 1–33). Chauvet, L. & Mesplé-Somps, S. 2007. Impact des financements internationaux sur les inégalités des pays en développement. Revue économique, 58(2007/3), 735–744. Edwards, A. C. & Ureta, M. 2003. International migration, remittances, and schooling: evidence from el salvador. Journal of Development Economics, 72(2), 429–461.

Falkingham, J., 2004, Poverty, out-of-pocket payments and access to health care: evidence fromTajikistan, Social Science & Medicine 58, 247–258

Fajnzylber, P. & Lopez, J. 2007. Close to Home the Development Impact of Remittances in Latin

America. the International Bank for Reconstruction and Development. Fay, M., Leipziger, D., Wodon, Q. & Yepes, T., 2005. “Achieving Child-Health-Related Millenium Development Goals: The Role of Infrastructure”, World Development 33(8): 1267-1284. Filmer, D. & Pritchett, L. 1999. The impact of public spending on health: does money matter? Social

Science & Medicine, 49(10), 1309–1323. Freund, C. & Spatafora, N. (2008). Remittances, transaction costs, and informality. Journal of

Development Economics, 86(2), 356–366. Gage, A. J., 2007, Barriers to the utilization of maternal health care in rural Mali, Social Science &

Medicine 65, 1666–1682


26

Gauthier, B. & Wane, W. 2009. Leakage of public resources in the health sector: An empirical investigation of chad. Journal of African Economies, 18(1), 52–83. Gupta, S., Pattillo, C. A., & Wagh, S. 2009. Effect of remittances on poverty and financial development in sub-saharan africa. World Development, 37(1), 104–115. Gwatkin, D. 2003. Reducing health inequalities in developing countries. accessed online at http://poverty. worldbank. org/files/9219_OTPHPaper-Gwatkin. pdf, on June, 5. Gwatkin, D., Rutstein, S., Johnson, K., Suliman, E., Wagstaff, A., & Amouzou, A. 2007. Socio-economic differences in health, nutrition, and population within developing countries. Hanson, G. & Woodruff, C. 2003. Emigration and educational attainment in Mexico. University of California at San Diego. Mimeographed. Hildebrandt, N. & McKenzie, D. 2005. The effects of migration on child health in Mexico. Policy Research Working Paper Series 3573, The World Bank. Kamgnia D. B. 2000. Distribution Impact of Public Spending in Cameroon: The Case of Health Care, AERC Research Paper 179, May. Kamgnia, D. B., Wangun, S. L., Tatsinkou, C., & Afor, J. 2008. Bénéfices acquis et ciblage des pauvres dans les dépenses publiques de santé et d’éducation au Cameroun. Cahiers de recherche PMMA 2008-08, PEP-PMMA. Kanaiaupuni, S. & Donato, K. 1999. Migradollars and mortality: The effects of migration on infant survival in Mexico. Demography, (pp. 339–353). Koechlin, V. & Leon, G. 2007. International Remittances and Income Inequality: An Empirical Investigation. Journal of Policy Reform, 10(2). Lindstrom, D. & Munoz-Franco, E. 2006. Migration and maternal health services utilization in rural Guatemala. Social Science & Medicine, 63(3), 706–721. Lopez-Cordova, E. 2004. Globalization, Migration and Development: The Role of Mexican Migrant Remittances. Economia, 6(1). Murawski, L. and Church, R. L., 2009, Improving accessibility to rural health services: The maximal covering network improvement problem, Socio-Economic Planning Sciences 43 (2009) 102–110 Rajkumar, A. & Swaroop, V. 2008. Public spending and outcomes: Does governance matter? Journal of Development Economics, 86(1), 96–111. Ratha, D. & Zhimei, X. 2008. Migration and remittances factbook 2008. World Bank Publications. Ravallion, M., 2007. Achieving Child-Health-Related Millenium Development Goals: The Role of Infrastructure – A Comment, World Development 35(5): 920-928. Reanne, F., Palma-Coca, O., Rauda-Esquivel, J., Olaiz-Fernandez, G., Diaz-Olavarrieta, C., & Acevedo-Garcia, D. 2009. The Relationship Between Remittances and Health Care Provision in Mexico. American Journal of Public Health, 99(7), 1227.


27

Reinikka, R. & Svensson, J. 2004. Local capture: Evidence from a central government transfer program in uganda. The Quarterly Journal of Economics, 119(2), 678–704. Reinikka, R. & Svensson, J. 2005. Fighting corruption to improve schooling: Evidence from a newspaper campaign in Uganda. Journal of the European Economic Association, 3(2-3), 259–267. Sepehri A., Sarma S., Simpson W. and Moshiri S., 2008, How important are individual, household and commune characteristics in explaining utilization of maternal health services in Vietnam?, Social

Science & Medicine 67, 1009–1017 Ssengonzi, R., De Jong, G., & Shannon Stokes, C. 2002. The effect of female migration on infant and child survival in Uganda. Population Research and Policy Review, 21(5), 403–431. Stock, J., Wright, J., & Yogo, M. 2002. A survey of weak instruments and weak identification in generalized method of moments. Journal of Business and Economic Statistics, 20(4), 518–529. Tavares, J. 2003. Does foreign aid corrupt? Economics Letters, 79(1), 99–106. Weil, D. 2007. Accounting for the Effect of Health on Economic Growth. The Quaterly Journal of

Economics, vol. 122, N°3, 1265- 1306 Woodruff, C. & Zenteno, R. 2007. Migration networks and microenterprises in mexico. Journal of

Development Economics, 82(2), 509–528. Yang, D. 2008. International migration, remittances and household investment: Evidence from Philippine migrants’ exchange rate shocks. Economic Journal, 118(528), 591–630. Yazbeck, A. 2009. Attacking Inequality in the Health Sector: A Synthesis of Evidence and Tools. Zachariah, K., Mathew, E., & Rajan, S. 2001. Impact of migration on Kerala’s economy and society. International Migration, 39(1), 63–87. Zweifel, P., Breyer, F., & Kifmann, F. (1997). Health economics, Oxford University Press.


28

APPENDIX

Table 1: DHS Data bases for selected developing countries

F. SUB-SAHARAN AFRICA

Benin 1996 Mali 1995/96

Benin 2001 Mali 2001

Burkina Faso 1992/93 Mauritania 2000/01

Burkina Faso 1998/99 Mozambique 1997

Burkina Faso 2003 Mozambique 2003

Cameroon 1991 Namibia 1992

Cameroon 1998 Namibia 2000

Cameroon 2004 Niger 1998

Central African Republic 1994/95 Nigeria 1990

Chad 1996/97 Nigeria 2003

Chad 2004 Rwanda 2000

Comoros 1996 Senegal 1997

Côte d'Ivoire 1994 South Africa 1998

Eritrea 1995 Tanzania 1996

A. EAST ASIA, PACIFIC B. EUROPE, CENTRAL ASIA

Cambodia 2000 Armenia 2000

Indonesia 1997 Kazakhstan 1995

Indonesia 2002/03 Kazakhstan 1999

Philippines 1998 Kyrgyzstan 1997

Philippines 2003 Turkey 1993

Vietnam 1997 Turkey 1998

Vietnam 2002 Turkmenistan 2000

Uzbekistan 1996

C. LATIN AMERICA, CARIBBEAN D. MIDDLE EAST, NORTH AFRICA

Bolivia 1998 Egypt 1995

Bolivia 2003 Egypt 2000

Brazil 1996 Jordan 1997

Colombia 1995 Morocco 1992

Colombia 2000 Morocco 2003/04

Colombia 2005 Yemen 1997

Dominican Republic 1996 E. SOUTH ASIA

Dominican Republic 2002 Bangladesh 1996/97

Guatemala 1995 Bangladesh 1999/2000

Guatemala 1998/99 Bangladesh 2004

Haiti 1994/95 India 1992/93

Haiti 2000 India 1998/99

Nicaragua 1997/98 Nepal 1996

Nicaragua 2001 Nepal 2001

Paraguay 1990 Pakistan 1990/91

Peru 1996


29

Ethiopia 2000 Tanzania 1999

Gabon 2000 Tanzania 2004

Ghana 1993 Togo 1998

Ghana 1998 Uganda 1995

Ghana 2003 Uganda 2000/01

Guinea 1999 Zambia 1996

Kenya 1993 Zambia 2001

Kenya 1998 Zimbabwe 1994

Kenya 2003 Zimbabwe 1999

Madagascar 1997

Malawi 1992

Malawi 2000

Source: Gwatkin et al. (2007)

Table 2: Descriptive statistics

Variables Obs Mean Std. Dev. Min Max

Fever 375 3.537339 0.5315214 1.791759 4.437934

Fever_pub 375 3.074057 0.6076348 1.223776 4.200205

Fever_priv 373 2.088516 1.074513 0.0000 4.283587

Diarrhea 392 3.436393 0.4820075 2.04122 4.412798

Diarrhea_pub 392 3.002404 0.5749169 0.6931472 4.309456

Diarrhea_priv 383 1.978827 1.039611 0.0000 4.207673

GDP per capita (log) 392 6.398928 0.8994376 4.796733 8.36346

Remittances per capita (log) 392 2.455868 1.365903 0.0825321 6.013539

Health aid per capita (log) 392 1.076897 0.5323376 0.1076645 2.218926

Public health spending per capita (log) 255 2.538217 1.153726 0.278582 4.913576

Mother education (log) 392 3.589526 1.012166 0.7419373 4.612146


30

Table A1: Instrumentation equations for remittances and health aid per capita

Model of fever Model of fever in public facilities

Model of fever in private facilities

Model of diarrhea Model of diarrhea in public facilities

Model of diarrhea in public facilities

Dependent Variable Rem Aid Rem Aid Rem Aid Rem Aid Rem Aid Rem Aid

1 2 3 4 5 6 7 8 9 10 11 12

Included instruments

GDP per capita 3.904*** 0.675 3.904*** 0.675 3.893*** 0.760 4.221*** 0.858* 4.221*** 0.858* 3.933*** 0.951**

(3.94) (1.37) (3.94) (1.37) (3.90) (1.55) (4.52) (1.85) (4.52) (1.85) (4.12) (2.04)

GDP per capita² -0.230*** -0.051 -0.230*** -0.051 -0.230*** -0.057 -0.255*** -0.067* -0.255*** -0.067* -0.234*** -0.073**

(3.05) (1.36) (3.05) (1.36) (3.01) (1.53) (3.58) (1.90) (3.58) (1.90) (3.21) (2.07)

Mother Education -0.316*** 0.040 -0.316*** 0.040 -0.307*** 0.038 -0.351*** 0.034 -0.351*** 0.034 -0.341*** 0.020

(5.62) (1.32) (5.62) (1.32) (5.31) (1.21) (6.48) (1.16) (6.48) (1.16) (6.01) (0.65)

Quintiles_2 0.135 -0.013 0.135 -0.013 0.124 -0.010 0.125 -0.002 0.125 -0.002 0.128 -0.008

(0.90) (0.16) (0.90) (0.16) (0.83) (0.12) (0.85) (0.02) (0.85) (0.02) (0.86) (0.09)

Quintiles_3 0.230 -0.031 0.230 -0.031 0.216 -0.027 0.220 -0.020 0.220 -0.020 0.240 -0.014

(1.50) (0.36) (1.50) (0.36) (1.39) (0.31) (1.45) (0.24) (1.45) (0.24) (1.55) (0.17)

Quintiles_4 0.336** -0.056 0.336** -0.056 0.321* -0.040 0.329** -0.038 0.329** -0.038 0.348** -0.018

(2.07) (0.63) (2.07) (0.63) (1.94) (0.45) (2.06) (0.44) (2.06) (0.44) (2.12) (0.21)

Quintiles_5 0.485*** -0.058 0.485*** -0.058 0.465** -0.052 0.477*** -0.042 0.477*** -0.042 0.493*** -0.027

(2.70) (0.61) (2.70) (0.61) (2.53) (0.53) (2.70) (0.45) (2.70) (0.45) (2.71) (0.28)

Constant -14.240*** -1.425 -14.240*** -1.425 -14.138*** -1.649 -15.255*** -1.998 -15.255*** -1.998 -14.292*** -2.332

(4.67) (0.92) (4.67) (0.92) (4.61) (1.06) (5.33) (1.37) (5.33) (1.37) (4.92) (1.58)

Excluded instruments

Financial development 0.856*** -0.047 0.856*** -0.047 0.843*** -0.062 0.878*** -0.031 0.878*** -0.031 0.881*** -0.031

(8.50) (0.86) (8.50) (0.86) (8.09) (1.14) (9.05) (0.61) (9.05) (0.61) (8.70) (0.59)

Aid weighted by religion -0.025** 0.022*** -0.025** 0.022*** -0.025** 0.023*** -0.027** 0.022*** -0.027** 0.022*** -0.032*** 0.024***

(2.19) (3.74) (2.19) (3.74) (2.24) (3.77) (2.52) (3.87) (2.52) (3.87) (2.91) (4.01)

Aid weighted by language -0.058*** 0.031*** -0.058*** 0.031*** -0.060*** 0.029*** -0.052*** 0.032*** -0.052*** 0.032*** -0.053*** 0.034***

(3.45) (4.33) (3.45) (4.33) (3.46) (4.11) (3.25) (4.93) (3.25) (4.93) (3.06) (5.07)

Observations 380 380 380 380 378 378 392 392 392 392 383 383


31

Table A2: Instrumentation equations for remittances and public health spending per capita

Model of fever Model of fever in public facilities

Model of fever in private facilities

Model of diarrhea Model of diarrhea in public facilities

Model of diarrhea in public facilities

Dependent Variable Rem Health Rem Health Rem Health Rem Health Rem Health Rem Health

1 2 3 4 5 6 7 8 9 10 11 12

Included instruments

GDP per capita 1.531 -1.500** 1.531 -1.500** 1.531 -1.500** 2.082* -1.848*** 2.082* -1.848*** 1.925 -1.900***

(1.18) (2.11) (1.18) (2.11) (1.18) (2.11) (1.71) (2.95) (1.71) (2.95) (1.54) (2.95)

GDP per capit² -0.053 0.189*** -0.053 0.189*** -0.053 0.189*** -0.103 0.219*** -0.103 0.219*** -0.087 0.224***

(0.52) (3.50) (0.52) (3.50) (0.52) (3.50) (1.08) (4.71) (1.08) (4.71) (0.89) (4.69)

Mother educ. -0.209*** 0.130*** -0.209*** 0.130*** -0.209*** 0.130*** -0.235*** 0.124*** -0.235*** 0.124*** -0.275*** 0.112***

(2.87) (3.34) (2.87) (3.34) (2.87) (3.34) (3.23) (3.18) (3.23) (3.18) (3.87) (2.80)

Quintiles_2 0.090 -0.049 0.090 -0.049 0.090 -0.049 0.085 -0.045 0.085 -0.045 0.101 -0.041

(0.47) (0.57) (0.47) (0.57) (0.47) (0.57) (0.45) (0.55) (0.45) (0.55) (0.54) (0.50)

Quintiles_3 0.139 -0.086 0.139 -0.086 0.139 -0.086 0.126 -0.091 0.126 -0.091 0.182 -0.074

(0.70) (0.97) (0.70) (0.97) (0.70) (0.97) (0.64) (1.06) (0.64) (1.06) (0.93) (0.85)

Quintiles_4 0.205 -0.128 0.205 -0.128 0.205 -0.128 0.177 -0.127 0.177 -0.127 0.245 -0.106

(0.97) (1.35) (0.97) (1.35) (0.97) (1.35) (0.84) (1.40) (0.84) (1.40) (1.17) (1.15)

Quintiles_5 0.311 -0.187* 0.311 -0.187* 0.311 -0.187* 0.277 -0.180* 0.277 -0.180* 0.362 -0.153

(1.29) (1.78) (1.29) (1.78) (1.29) (1.78) (1.16) (1.81) (1.16) (1.81) (1.52) (1.52)

Constant -6.972* 3.491 -6.972* 3.491 -6.972* 3.491 -8.383** 4.445** -8.383** 4.445** -8.073** 4.547**

(1.75) (1.59) (1.75) (1.59) (1.75) (1.59) (2.23) (2.26) (2.23) (2.26) (2.11) (2.26)

Excluded instruments

Fin. Dev. 0.632*** -0.059 0.632*** -0.059 0.632*** -0.059 0.581*** -0.015 0.581*** -0.015 0.654*** 0.009

(4.44) (1.04) (4.44) (1.04) (4.44) (1.04) (4.25) (0.28) (4.25) (0.28) (4.76) (0.16)

HSN -0.081 0.414*** -0.081 0.414*** -0.081 0.414*** -0.051 0.407*** -0.051 0.407*** -0.050 0.407***

(0.65) (12.47) (0.65) (12.47) (0.65) (12.47) (0.40) (12.33) (0.40) (12.33) (0.40) (12.24)

MSP 0.207 -0.262*** 0.207 -0.262*** 0.207 -0.262*** 0.244 -0.285*** 0.244 -0.285*** 0.205 -0.297***

(1.44) (6.86) (1.44) (6.86) (1.44) (6.86) (1.65) (7.57) (1.65) (7.57) (1.38) (7.68)

Observations 218 218 218 218 218 218 225 225 225 225 223 223

Note: HSN and MSP are average public sector health spending per capita and the average defense spending per capita of a country’s geographic neighbors respectively.